Bing 3
by DerekZahn
Fri Jan 05, 2007 8:25 am
Hey Jon!
I've been watching your progress, it looks great so far. I think our approaches are somewhat similar.
The sensors haven't gone to work yet in the real world but on the workbench they seem great. I haven't really described exactly how I'll be using them, but this diagram should illustrate the idea:
http://happyrobots.com/footpads.gif
I've been watching your progress, it looks great so far. I think our approaches are somewhat similar.
The sensors haven't gone to work yet in the real world but on the workbench they seem great. I haven't really described exactly how I'll be using them, but this diagram should illustrate the idea:
http://happyrobots.com/footpads.gif
by bauermech
Fri Jan 05, 2007 6:17 pm
For the same reason you don't have 40lb feet yourself -- the more maneuverable the legs are the better.
Yeah, I noticed how much strain is put on the hip servos when Rook's Pawn (v1) lifted his leg. His movements had to be dramatically slowed down or he'd loose his balance on one leg. The extra weight is nice when he was standing in place, but kinda' defetes the purpose of having a walking bi-ped robot.
I'm looking for that happy medium between stability and agility.
by DerekZahn
Sat Jan 06, 2007 10:33 pm
Some sideways progress: the feet have been rebuilt. Now to move upwards again!
bigger: http://happyrobots.com/bing3i2.jpg
bigger: http://happyrobots.com/bing3i2.jpg
by DerekZahn
Mon Jan 08, 2007 4:27 am
Bing's legs are finally united!
bigger: http://happyrobots.com/bing3j2.jpg
The picture is a little bit of a cheat because it implies that the lower torso section is done, but it really isn't. That's the next task, and it will probably take all week so there won't be more photos for a little while as I work on various details.
bigger: http://happyrobots.com/bing3j2.jpg
The picture is a little bit of a cheat because it implies that the lower torso section is done, but it really isn't. That's the next task, and it will probably take all week so there won't be more photos for a little while as I work on various details.
by DerekZahn
Mon Jan 08, 2007 6:20 pm
Thanks for the compliments!
The height is on target to be 32cm to the top of the shoulders, I'm not sure exactly what the height is at the moment, probably about 25 cm. I was hoping that the total weight would end up about 2.8 kg, but I think 3.0 kg is more likely and it could even end up a bit heavier. I don't have an appropriate scale at the moment to get an exact weight.
The blue parts are 1.5mm (0.06 inch) aluminum sheet, painted blue.
The height is on target to be 32cm to the top of the shoulders, I'm not sure exactly what the height is at the moment, probably about 25 cm. I was hoping that the total weight would end up about 2.8 kg, but I think 3.0 kg is more likely and it could even end up a bit heavier. I don't have an appropriate scale at the moment to get an exact weight.
The blue parts are 1.5mm (0.06 inch) aluminum sheet, painted blue.
by Robo1
Mon Jan 08, 2007 6:45 pm
What controller are you going to use for this bot.
have you considered using an FPGA.
you could have all the servos constantly updated and the pot measured with out using interrupts. and a master controller updating the FPGA with the positions. This is the method I'm building for a replacement board for my KHR-2.
bren
have you considered using an FPGA.
you could have all the servos constantly updated and the pot measured with out using interrupts. and a master controller updating the FPGA with the positions. This is the method I'm building for a replacement board for my KHR-2.
bren
by DerekZahn
Mon Jan 08, 2007 7:42 pm
Bing is controlled via a serial bus, so the main central controller is anything with a serial port. In the CAD I'm finishing up, I just have a cavity for this controller, but I haven't decided exactly what to use. I want the most powerful very small controller I can get. Presently I'm planning on using a 400 mhz gumstix, but that could change. For the first couple months of software development I will be writing code on a PC and running the robot off the PC serial port, because it's easy for me to write code in that environment. Then later I'll recompile the controller code for whatever onboard controller I end up using.
Your approach should work great too! Keep us posted!
Your approach should work great too! Keep us posted!
by DerekZahn
Thu Jan 11, 2007 5:58 pm
Most of the wiring for the lower torso is complete, and I thought I'd describe Bing's communication scheme, which is in some ways a bit peculiar.
The processor module that I'm using for various tasks (the NewMicros tini2131) has a built-in RS232 level converter, so RS232 is the communication mechanism. Each processor controls 3 servos (and may have additional sensors), so since Bing has 21 servos, that's 7 processor modules. One additional module is the one for accelerometers and gyros, for a total of 8 processors. These processors are daisy-chained together so that the TX from one module goes to the RX of the next. In this way the only wiring that needs to be routed from one point to another is +V,GND,RX,TX.
The protocol is designed for low latency -- The main controller sends out servo positions in a big packet (about 50 bytes total), but each node in the chain sends a byte as soon as it gets a byte -- it doesn't need to wait for the whole packet. If it sees a servo position command intended for one of the servos it is controlling, it replaces those bytes with the bytes for a position report command so when the packet gets back to the main controller, it contains servo positions instead of the original servo commands. If a module wants to add additional sensors (like the pressure sensors in the foot), it appends them to the end of the packet.
So far I have 6 of the 8 modules chained together (just the ones for the arms are not done yet). A complete cycle from the time that the PC queues up the packet until the whole packet is received back again from Bing is about 11 milliseconds, which is plenty fast for my needs.
I hope that made sense. It might seem odd to use RS232 in this way -- other protocols like SPI or I2C are possibly better, but I know how to use RS232 and the pins are sitting there anyway so I thought I'd use them.
The processor module that I'm using for various tasks (the NewMicros tini2131) has a built-in RS232 level converter, so RS232 is the communication mechanism. Each processor controls 3 servos (and may have additional sensors), so since Bing has 21 servos, that's 7 processor modules. One additional module is the one for accelerometers and gyros, for a total of 8 processors. These processors are daisy-chained together so that the TX from one module goes to the RX of the next. In this way the only wiring that needs to be routed from one point to another is +V,GND,RX,TX.
The protocol is designed for low latency -- The main controller sends out servo positions in a big packet (about 50 bytes total), but each node in the chain sends a byte as soon as it gets a byte -- it doesn't need to wait for the whole packet. If it sees a servo position command intended for one of the servos it is controlling, it replaces those bytes with the bytes for a position report command so when the packet gets back to the main controller, it contains servo positions instead of the original servo commands. If a module wants to add additional sensors (like the pressure sensors in the foot), it appends them to the end of the packet.
So far I have 6 of the 8 modules chained together (just the ones for the arms are not done yet). A complete cycle from the time that the PC queues up the packet until the whole packet is received back again from Bing is about 11 milliseconds, which is plenty fast for my needs.
I hope that made sense. It might seem odd to use RS232 in this way -- other protocols like SPI or I2C are possibly better, but I know how to use RS232 and the pins are sitting there anyway so I thought I'd use them.
by Robo1
Thu Jan 11, 2007 6:16 pm
Sounds good to me.
Hows the foot pressure sensors coming on. I've been looking in to designing the feet similar to your's but with little distance sensors all around the base so that hopefully it can climb stairs automatically. I will also have some gyros in the foot to.
All this will hopefully make the ankle fully autonomose so that I don't have to worry about coding the position unless you want them in certain positions like when they walking. I also hope this will help in uneven terrain as the foot wil auto position it self.
bren
Hows the foot pressure sensors coming on. I've been looking in to designing the feet similar to your's but with little distance sensors all around the base so that hopefully it can climb stairs automatically. I will also have some gyros in the foot to.
All this will hopefully make the ankle fully autonomose so that I don't have to worry about coding the position unless you want them in certain positions like when they walking. I also hope this will help in uneven terrain as the foot wil auto position it self.
bren
by DerekZahn
Thu Jan 11, 2007 6:26 pm
Those distance sensors sound like they would be very useful!
The pressure sensors seem to be fine... when I push on them the numbers in my little PC test app change and the numbers seem roughly linear with applied pressure (up to a limit of about 2 kg where the sensor saturates -- that limit is determined by my circuitry, not by the sensor itself). I won't start on the "hard part" -- smart control software -- until the construction is complete, so I don't know more than that at this point.
I said in an earlier message that you could get these sensors at phidgets.com, but they appear not to sell them any more and they are now carried by www.trossenrobotics.com
The pressure sensors seem to be fine... when I push on them the numbers in my little PC test app change and the numbers seem roughly linear with applied pressure (up to a limit of about 2 kg where the sensor saturates -- that limit is determined by my circuitry, not by the sensor itself). I won't start on the "hard part" -- smart control software -- until the construction is complete, so I don't know more than that at this point.
I said in an earlier message that you could get these sensors at phidgets.com, but they appear not to sell them any more and they are now carried by www.trossenrobotics.com
by DerekZahn
Fri Jan 12, 2007 12:56 am
The lower part of the torso is finished, non-fake photo:
bigger: http://happyrobots.com/bing3k2.jpg
Here's a screenshot of the little shell PC program that I use to talk to Bing to make sure things are working right.
bigger: http://happyrobots.com/bing3l2.gif
Next I will start working on the arms and upper torso (not sure which to work on first). Because I am most interested in walking, I will be sticking with simple 3-servo arms for now, using some JR servos I got a long time ago.
bigger: http://happyrobots.com/bing3k2.jpg
Here's a screenshot of the little shell PC program that I use to talk to Bing to make sure things are working right.
bigger: http://happyrobots.com/bing3l2.gif
Next I will start working on the arms and upper torso (not sure which to work on first). Because I am most interested in walking, I will be sticking with simple 3-servo arms for now, using some JR servos I got a long time ago.